Answer variability
Answer variability is the way an AI engine can give different answers to the same prompt when it is asked more than once, including naming different brands or citing different sources.
Example
You ask an AI engine the same question about accounting software on three different days. The first answer lists your brand first, the second lists it third, and the third leaves it out entirely.
Why it happens
Several factors can play a part: the model generates text with some randomness, search results used for grounding can change, models are updated, and some engines tailor answers to location or conversation history. The mix differs by engine and is rarely documented in detail.
Why it matters
A single screenshot of an AI answer is one draw from a range of possible answers. Decisions based on one run can overreact to noise. Running prompts repeatedly and reporting a range, such as a P25-P75 range, shows both the typical result and how settled it is.
How Stellarcast handles it
Stellarcast re-runs tracked prompts on a schedule and reports visibility and share of voice with a P25-P75 range that shows how much each figure could still move as more days of readings come in. A lift from a fix is labeled MEASURED only when the change sits outside the range observed before the fix.
See how AI describes your brand
Get a free AI visibility audit across ChatGPT, Gemini, Perplexity and Google AI Overviews and see where you stand against your competitors.
Get your free visibility audit